U.S. patent application number 11/417769 was filed with the patent office on 2007-07-19 for magnetic resonance spatial risk map for tissue outcome prediction.
Invention is credited to Nina Menezes, Alma Gregory Sorensen.
Application Number | 20070167727 11/417769 |
Document ID | / |
Family ID | 38264112 |
Filed Date | 2007-07-19 |
United States Patent
Application |
20070167727 |
Kind Code |
A1 |
Menezes; Nina ; et
al. |
July 19, 2007 |
Magnetic resonance spatial risk map for tissue outcome
prediction
Abstract
Diffusion weighted images and perfusion weighted images are
acquired with an MRI system and used to produce low b, DWI, ADC,
CBV, CBF, and MTT images of brain tissues following an ischemic
event. These MRI physiological measurements are input along with a
spatial location measurement to a generalized linear model that
predicts the outcome of tissues surrounding a lesion.
Inventors: |
Menezes; Nina; (Boston,
MA) ; Sorensen; Alma Gregory; (Lexington,
MA) |
Correspondence
Address: |
QUARLES & BRADY LLP
411 E. WISCONSIN AVENUE
SUITE 2040
MILWAUKEE
WI
53202-4497
US
|
Family ID: |
38264112 |
Appl. No.: |
11/417769 |
Filed: |
May 4, 2006 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60678434 |
May 6, 2005 |
|
|
|
Current U.S.
Class: |
600/410 |
Current CPC
Class: |
G06T 2207/30016
20130101; G06T 7/0012 20130101; G01R 33/56341 20130101 |
Class at
Publication: |
600/410 |
International
Class: |
A61B 5/05 20060101
A61B005/05 |
Goverment Interests
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH
[0002] This invention was made with government support under Grant
No. 5R01NS038477-07 awarded by the National Institute of Health.
The United States Government has certain rights in this invention.
Claims
1. A method for predicting tissue fate, the steps comprising: a)
acquiring data from the tissues with a magnetic resonance imaging
(MRI) system that contains information indicative of a
physiological parameter related to tissue health; b) reconstructing
an image of the tissues from the acquired data; c) calculating from
the acquired data a physiological parameter at tissue locations in
the reconstructed image; d) selecting a core region in the
reconstructed image; e) calculating a location parameter related to
the distance of each of said tissue locations from the core region;
and f) predicting the fate of tissues at each of said tissue
locations using a predetermined model that employs as inputs the
physiological parameter and the location parameter.
2. The method as recited in claim 1 in which the predetermined
model is a generalized linear model of the form: P = 1 1 + e -
.alpha. + .beta. .times. .times. x + .gamma. .times. .times. r
##EQU4## in which P is the predicted outcome, x is the
physiological parameter, r is the location parameter, .beta. is a
coefficient that weights the physiological parameter, .gamma. is a
coefficient that weights the location parameter, and .alpha. is a
bias term.
3. The method as recited in claim 1 in which the location parameter
is equal to the distance of the tissue location from the core
region.
4. The method as recited in claim 1 in which the tissues are
located in a mammalian brain.
5. The method as recited in claim 1 in which step a) includes
acquiring a diffusion weighted image.
6. The method as recited in claim 5 in which a plurality of
diffusion weighted images are acquired with motion encoding
gradients oriented in respective different directions.
7. The method as recited in claim 6 in which the physiological
parameter calculated in step c) includes apparent diffusion
coefficient (ADC) calculated from the plurality of diffusion
weighted images.
8. The method as recited in claim 1 in which step a) includes
acquiring a set of perfusion weighted image frames.
9. The method as recited in claim 8 in which step c) includes
calculating a plurality of physiological parameters from said set
of perfusion weighted image frames, and in which the predetermined
model accepts said plurality of physiological parameters as inputs
to the prediction.
10. The method as recited in claim 9 in which step a) includes
acquiring a diffusion weighted image, and in which step c) includes
calculating a diffusion coefficient at each tissue location from
the acquired diffusion weighted image.
11. The method as recited in claim 1 in which step d) is performed
by manually selecting the core region in the image reconstructed in
step b).
12. The method as recited in claim 1 in which step d) is performed
by selecting tissue locations based on the physiological parameter
values calculated in step c).
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is based on U.S. Provisional patent
application Ser. No. 60/678,434 filed on May 6, 2005 and entitled
"Method For Predicting Tissue Outcome in Acute Human Stroke."
BACKGROUND OF THE INVENTION
[0003] The field of the invention is nuclear magnetic resonance
imaging (MRI) methods and systems. More particularly, the invention
relates to MR imaging of the brain.
[0004] When a substance such as human tissue is subjected to a
uniform magnetic field (polarizing field B.sub.0), the individual
magnetic moments of the spins in the tissue attempt to align with
this polarizing field, but precess about it in random order at
their characteristic Larmor frequency. If the substance, or tissue,
is subjected to a magnetic field (excitation field B.sub.1) which
is in the x-y plane and which is near the Larmor frequency, the net
aligned moment, Mz, may be rotated, or "tipped", into the x-y plane
to produce a net transverse magnetic moment Mt. A signal is emitted
by the excited spins after the excitation signal B.sub.1 is
terminated, this signal may be received and processed to form an
image.
[0005] When utilizing these signals to produce images, magnetic
field gradients (G.sub.x, G.sub.y and G.sub.z) are employed.
Typically, the region to be imaged is scanned by a sequence of
measurement cycles in which these gradients vary according to the
particular localization method being used. The resulting set of
received NMR signals are digitized and processed to reconstruct the
image using one of many well known reconstruction techniques.
[0006] Magnetic resonance imaging currently plays an essential role
in the diagnosis of stroke, both in distinguishing between
hemorrhage and ischemia, and in determining the extent and
localization of the lesion. Another important goal for patient
management is prognosis. In terms of clinical decision-making
regarding therapeutic options, the challenge of providing early
prognosis in stroke can be broken down into two parts. First, how
likely is the ischemic tissue to infarct in the absence of
intervention? This is a problem of predicting tissue outcome.
Second, if the ischemic tissue does infarct, how critical will the
resulting cognitive and behavioral deficit be? This is problem of
predicting clinical outcome. The goal, ultimately, is to improve
stroke patient care, and to accomplish this will require the
accurate prediction of tissue fate and the means to translate the
prediction of tissue fate into one of clinical fate. Tools that
accurately estimate both tissue and clinical outcome in the acute
setting would dramatically impact patient care. For example,
patients identified early on as having a good prognosis can be
spared risky therapeutics. Conversely, early identification of poor
prognosis will weight heavily in the decision of whether to use
thrombolytics that carry a certain amount of risk.
[0007] There are many MR imaging techniques used to acquire
diagnostic information from the brain. These include contrast
enhanced T.sub.1-weighted images that brightly reveal regions where
the blood-brain barrier is destroyed, T.sub.2-weighted
fast-spin-echo (FSE) and fluid attenuated inversion-recovery
(FLAIR) imaging which show the extent of edema surrounding a
damaged region. Two of the most important diagnostic tools,
however, are diffusion-weighted imaging (DWI) and
perfusion-weighted imaging (PWI) which measure physiological
parameters that correlate with tissue health.
[0008] Diffusion-weighted imaging (DWI) is a powerful MRI technique
for probing microscopic tissue structure. In DWI, a pulse sequence
is employed which contains a magnetic field gradient known as a
diffusion gradient that sensitizes the MR signal to spin motion. In
a DWI pulse sequence the detected MR signal intensity decreases
with the speed of water diffusion in a given volume of tissue. The
first moment of this diffusion gradient, also known as the
"b-value" determines the speed of diffusion to which the image is
sensitive. This b-value may be adjusted by either varying the area
of the two lobes of the diffusion magnetic field gradient, or by
varying the time interval between them. When water motion in the
subject is unrestricted, the MR signal intensity at the center of
the echo using a spin-echo diffusion-weighted pulse sequence is
related to the b-value as follows: A = S .function. ( b ) S 0 = e -
bD ( 1 ) ##EQU1## where the "b-value"
b=.gamma..sup.2G.sup.2.delta..sup.2(.DELTA.-.delta./3). The
parameter .gamma. is the gyromagnetic ratio and G is the amplitude
of the applied diffusion magnetic field gradients. S(b) is the MR
signal magnitude with diffusion weighting b, and S.sub.0 is the MR
signal magnitude with no diffusion weighting (b=0). The parameter D
is the diffusion coefficient of the fluid (in mm.sup.2/s), which
directly reflects the fluid viscosity where there are no structural
restrictions to diffusion of the water. .DELTA. is the time
interval between the onsets of the two diffusion gradient lobes and
.delta. is the duration of each gradient lobe. The diffusion
coefficient D in equation (1) may be calculated, since b is known
and the attenuation A can be measured.
[0009] The interpretation of attenuation A becomes complicated when
water molecules are restricted in their motion by tissue
structures. Different populations of water within a voxel then
diffuse, on average, at different rates. One can fit the measured
attenuation data with a mono-exponential function, or make an
estimate of the signal decay rate using a single b-value, yielding
an apparent diffusion coefficient (ADC). The ADC is useful, in
detecting cytotoxic edema following a stroke.
[0010] Perfusion as related to tissue refers to the exchange of
oxygen, water and nutrients between blood and tissue. The
measurement of tissue perfusion is important for the functional
assessment of organ health. Perfusion weighted images (PWI) which
show by their brightness the degree to which tissues are perfused
can be used, to assess the health of brain tissues that have been
damaged by a stroke. A number of methods have been used to produce
perfusion images using magnetic resonance imaging techniques. One
technique, as exemplified by U.S. Pat. No. 6,295,465, is to
determine the wash-in or wash-out kinetics of contrast agents such
as chelated gadolinium. From the acquired NMR data, images are
produced which indicate cerebral blood flow (CBF) at each voxel,
cerebral blood volume (CBV) at each voxel and mean transit time
(MTT) at each voxel. Each of these perfusion indication
measurements provides information that is useful in diagnosing
tissue health.
[0011] Several studies have noted that DWI- and PWI-derived
parameter values, such as the apparent diffusion coefficient (ADC)
and cerebral blood flow (CBF), vary on a voxel-by-voxel basis
within the ischemic territory, and their values have been found to
be associated with the likelihood of infarction. However, no single
parameter has been shown to be definitively predictive of
infarction, suggesting a multiparametric approach.
[0012] Models have been created to correlate the DWI and PWI
measurements to tissue outcome. One such method is described by Wu
et al "Predicting Tissue Outcome In Acute Human Cerebral Ischemia
Using Combined Diffusion- and Perfusion-Weighted MR Imaging",
Stroke, 2001; 32:933-942 and is referred to as the generalized
linear model (GLM). With this predictive strategy a model is
created that relates predicted outcome P (0=normal,1=infarcted) to
the DWI and PWI measurements with the logistic function: P = 1 1 +
e - .alpha. + .beta. .times. .times. x ( 2 ) ##EQU2## where:
.alpha.=bias or intercept term that provides the base value for P
if all the input parameter x are zero,
[0013] .beta.=a vector of the coefficients used to weight each DWI
and PWI parameter measurement,
[0014] x=the respective DWI and PWI parameter measurements at the
voxel.
[0015] The vector .beta. is derived from training data acquired
from previous patients where the outcomes are known. As described
in the above-cited publication and in co-pending U.S. patent
application Ser. No. 10/182,978 entitled "Method For Evaluating
Novel, Stroke Treatments Using A Tissue Risk Map" this includes
selecting training regions in follow-up exams of a stroke patient
population and manually selecting regions in T2 weighted images
that clearly depict infarcted and noninfarcted tissues. The values
from these regions in earlier acquired DWI and PWI parameter images
from these same patients were used as the input vector x in the
training step. The coefficients (.beta.) are calculated using an
iterative reweighted least-squares algorithm.
SUMMARY OF THE INVENTION
[0016] The present invention is an improved tissue outcome
predictive model and a method for using that model to predict the
outcome of ischemic tissue. More specifically, the improved model
includes a parameter that indicates the voxel location with respect
to the core area of the lesion. A limitation of prior models is
that they fail to take into consideration the location-dependent
vulnerability to infarction of a voxel. The location parameter may
include a distance between the voxel and a core lesion, the
location parameter may reflect the region in the brain the voxel is
located, where different regions demonstrate different
vulnerabilities to clinical deficits, or where the location
parameter may indicate the collateral blood supply.
[0017] A discovery of the present invention is that multiparametric
models can be significantly improved by taking into account
location-dependent vulnerability to infarction. With a GLM model,
for example, location parameters may be added as inputs: P = 1 1 +
e - .alpha. + .beta. .times. .times. x + .gamma. .times. .times. r
( 3 ) ##EQU3## where: .gamma.=a vector of location parameter
weighting coefficients; and
[0018] r=corresponding location parameter inputs.
[0019] Another discovery of the present invention is a location
parameter that significantly increases the predictive performance
of the model. More specifically, a distance parameter is calculated
for each voxel that indicates its distance from the acute core
lesion. As shown in FIG. 2, for example, an acute DWI region 1 is
identified as the core area from which the ischemic injury extends
with time. PWI parameters may define a penumbra region 2 which
surrounds the core region 1 and which predicts tissue outcome
according to prior methods. According to the present invention
tissue outcome is also predicted by its distance from the core
region 1. The voxel 3 for example, is located a distanced d.sub.1
from the core 1, whereas the voxel 4 is located a distance d.sub.2
from the core lesion 1. Because voxel 4 is much closer to the core
lesion 1 than voxel 3, the location parameter d will predict a
higher chance of infarction for point 4 than point 3.
[0020] A general object of the invention is to provide an MR
imaging method that improves the tissue outcome prediction of
ischemic tissue. It has been discovered that the acute region
identified by DWI imaging can, and usually does, differ from the
acute region identified by PWI imaging. One of the two regions will
be smaller than the other and this smaller region defines the core
region where both measurement techniques agree that infarction will
occur. At voxels located outside this core region outcome is less
certain and the present invention adds location parameters to the
model to improve the outcome prediction.
[0021] The foregoing and other objects and advantages of the
invention will appear from the following description. In the
description, reference is made to the accompanying drawings which
form a part hereof, and in which there is shown by way of
illustration a preferred embodiment of the invention. Such
embodiment does not necessarily represent the full scope of the
invention, however, and reference is made therefore to the claims
and herein for interpreting the scope of the invention.
BRIEF DESCRIPTION OF THE DRAWINGS
[0022] FIG. 1 is a block diagram of an MRI system which employs the
present invention;
[0023] FIG. 2 is a schematic representation of the human brain
illustrating the spatial progression of ischemic injury;
[0024] FIG. 3 is a flow chart of the steps in a preferred method
for practicing the present invention; and
[0025] FIG. 4 is a flow chart of the steps used in the preferred
embodiment for producing the location-weighted map that forms part
of the method in FIG. 3
DESCRIPTION OF THE PREFERRED EMBODIMENT
[0026] Referring particularly to FIG. 1, the preferred embodiment
of the invention is employed in an MRI system. The MRI system
includes a workstation 10 having a display 12 and a keyboard 14.
The workstation 10 includes a processor 16 which is a commercially
available programmable machine running a commercially available
operating system. The workstation 10 provides the operator
interface which enables scan prescriptions to be entered into the
MRI system.
[0027] The workstation 10 is coupled to four servers: a pulse
sequence server 18; a data acquisition server 20; a data processing
server 22, and a data store server 23. In the preferred embodiment
the data store server 23 is performed by the workstation processor
16 and associated disc drive interface circuitry. The remaining
three servers 18, 20 and 22 are performed by separate processors
mounted in a single enclosure and interconnected using a 64-bit
backplane bus. The pulse sequence server 18 employs a commercially
available microprocessor and a commercially available quad
communication controller. The data acquisition server 20 and data
processing server 22 both employ the same commercially available
microprocessor and the data processing server 22 further includes
one or more array processors based on commercially available
parallel vector processors.
[0028] The workstation 10 and each processor for the servers 18, 20
and 22 are connected to a serial communications network. This
serial network conveys data that is downloaded to the servers 18,
20 and 22 from the workstation 10 and it conveys tag data that is
communicated between the servers and between the workstation and
the servers. In addition, a high speed data link is provided
between the data processing server 22 and the workstation 10 in
order to convey image data to the data store server 23.
[0029] The pulse sequence server 18 functions in response to
program elements downloaded from the workstation 10 to operate a
gradient system 24 and an RF system 26. Gradient waveforms
necessary to perform the prescribed scan are produced and applied
to the gradient system 24 which excites gradient coils in an
assembly 28 to produce the magnetic field gradients G.sub.x,
G.sub.y and G.sub.z used for position encoding NMR signals. The
gradient coil assembly 28 forms part of a magnet assembly 30 which
includes a polarizing magnet 32 and a whole-body RF coil 34.
[0030] RF excitation waveforms are applied to the RF coil 34 by the
RF system 26 to perform the prescribed magnetic resonance pulse
sequence. Responsive NMR signals detected by the RF coil 34 are
received by the RF system 26, amplified, demodulated, filtered and
digitized under direction of commands produced by the pulse
sequence server 18. The RF system 26 includes an RF transmitter for
producing a wide variety of RF pulses used in MR pulse sequences.
The RF transmitter is responsive to the scan prescription and
direction from the pulse sequence server 18 to produce RF pulses of
the desired frequency, phase and pulse amplitude waveform. The
generated RF pulses are applied to the whole body RF coil 34.
[0031] The RF system 26 also includes one or more RF receiver
channels. The RF receiver channel is connected to a receive coil,
which in the preferred embodiment is a head coil. The signal from
the head coil is coupled to an RF amplifier that amplifies the NMR
signal received by the coil, and a quadrature detector and
analog-to-digital converter detects and digitizes the I and Q
quadrature components of the received NMR signal. The magnitude of
the received NMR signal may thus be determined at any sampled
"k-space" point by the square root of the sum of the squares of the
I and Q components: M= {square root over (I.sup.2+Q.sup.2)}, (4)
and the phase of the received NMR signal may also be determined:
.phi.=tan.sup.-1Q/I. (5)
[0032] The pulse sequence server 18 also optionally receives
patient data from a physiological acquisition controller 36. The
controller 36 receives signals from a number of different sensors
connected to the patient, such as ECG signals from electrodes or
respiratory signals from a bellows. Such signals are typically used
by the pulse sequence server 18 to synchronize, or "gate", the
performance of the scan with the subject's respiration or heart
beat.
[0033] The pulse sequence server 18 also connects to a scan room
interface circuit 38 which receives signals from various sensors
associated with the condition of the patient and the magnet system.
It is also through the scan room interface circuit 38 that a
patient positioning system 40 receives commands to move the patient
to desired positions during the scan.
[0034] It should be apparent that the pulse sequence server 18
performs real-time control of MRI system elements during a scan. As
a result, it is necessary that its hardware elements be operated
with program instructions that are executed in a timely manner by
run-time programs. The description components for a scan
prescription are downloaded from the workstation 10 in the form of
objects. The pulse sequence server 18 contains programs which
receive these objects and converts them to objects that are
employed by the run-time programs.
[0035] The digitized NMR signal samples produced by the RF system
26 are received by the data acquisition server 20. The data
acquisition server 20 operates in response to description
components downloaded from the workstation 10 to receive the
real-time NMR data and provide buffer storage such that no data is
lost by data overrun. In some scans the data acquisition server 20
does little more than pass the acquired NMR data to the data
processor server 22. However, in scans which require information
derived from acquired NMR data to control the further performance
of the scan, the data acquisition server 20 is programmed to
produce such information and convey it to the pulse sequence server
18. For example, during prescans NMR data is acquired and used to
calibrate the pulse sequence performed by the pulse sequence server
18. Also, navigator signals may be acquired during a scan and used
to adjust RF or gradient system operating parameters or to control
the view order in which k-space is sampled. And, the data
acquisition server 20 may be employed to process NMR signals used
to detect the arrival of contrast agent in an MRA scan. In all
these examples the data acquisition server 20 acquires NMR data and
processes it in real-time to produce information which is used to
control the scan.
[0036] The data processing server 22 receives NMR data from the
data acquisition server 20 and processes it in accordance with
description components downloaded from the workstation 10. Such
processing may include, for example: Fourier transformation of raw
k-space NMR data to produce two or three-dimensional images; the
application of filters to a reconstructed image; the performance of
a backprojection image reconstruction of acquired NMR data; the
calculation of functional MR images; the calculation of motion or
flow images, etc.
[0037] Images reconstructed by the data processing server 22 are
conveyed back to the workstation 10 where they are stored.
Real-time images are stored in a data base memory cache (not shown)
from which they may be output to operator display 12 or a display
42 which is located near the magnet assembly 30 for use by
attending physicians. Batch mode images or selected real time
images are stored in a host database on disc storage 44. When such
images have been reconstructed and transferred to storage, the data
processing server 22 notifies the data store server 23 on the
workstation 10. The workstation 10 may be used by an operator to
archive the images, produce films, or send the images via a network
to other facilities.
[0038] Referring particularly to FIG. 3, the present invention is
practiced on the MRI system of FIG. 1. The first step as indicated
at process block 100 is to acquire NMR data from which a full
diffusion tensor image can be reconstructed. Each of the six
directions is acquired using a single-shot echo-planar (EPI) pulse
sequence with the first moment of the motion encoding gradient
waveform set to b=1000s/mm.sup.2 and then repeated without motion
encoding b=0. The EPI pulse sequence is a pulsed field gradient
spin-echo sequence with one motion encoding gradient lobe of d-47
ms disposed to one side of the 180.degree. RF refocusing pulse and
an identical motion encoding gradient lobe is disposed to the other
side of the refocusing pulse with a .DELTA.=52 ms spacing
therebetween. The TE is 118 ms and the TR is 6000. The phase
encoding is stepped to sample 128 lines of k-space and 256 k-space
samples are acquired from each line. A 256 by 128 array of DWI
k-space data is thus acquired. Each 2D slice has a thickness of 6
mm with a 1 mm interslice gap and a field of view of 40 cm by 20
cm. The number of acquired slices depends on the extent of the
brain to be examined, but typically 10 to 20 slices are acquired to
cover the volume to be measured. Reference is made to the following
publication for a more detailed description of the preferred DWI
acquisition: Sorensen A G, Buonanno FIRST AND SECOND, Gonzalez R G,
Schwamm L H, Lev M H, Huang-Hellinger F R, Reese T G, Weiskoff R M,
Davis T L, Suwanwela N, Can U, Moreira J A, Copen W A, Look R B,
Finklestein S P, Rosen B R, Koroshetz W J. Hyperacute Stroke:
Evaluation With Combined Multisection Diffusion-Weighted and
Hemodynamically Weighted Echo-Planar MR Imaging, Radiology, 1996;
199:391-401.
[0039] The next step as indicated at process block 102 is to
administer a contrast agent to the subject of the examination. A
0.2 mmol/kg of gadolinium-based contrast agent is injected at a
rate of 5 ml/s using an MRI-compatible power injector. The
injection is started 10s after commencing the PWI acquisition
described below and it is followed by a comparable volume of normal
saline at the same rate of 5 ml/s.
[0040] Perfusion-weighted imaging (PWI) is then performed as
indicated at process block 104 starting 10s before the contrast
injection. A time series of 2D images is acquired using a spin-echo
EPI pulse sequence. A total of 46 image frames are acquired from 10
to 20 slices during the first pass of the contrast agent. The size
and location of the slices is substantially the same as during the
DWI acquisition and PWI NMR data is acquired from the same volume
of voxels. A flip angle of 90.degree., a TE=75 ms and a TR=1.5 s
are used in the EPI pulse sequence. For a more detailed description
of the PWI acquisition reference is made to the following
publication: Sorensen A G, Copen W A, Ostergaard L, Buonanno FIRST
AND SECOND, Gonzalez R G, Rordorf G, Rosen B R, Schwamm L H,
Weisskoff R M, Koroshetz W J. Hyperacute Stroke: Simultaneous
Measurement Of Relative Cerebral Blood Volume. Relative Cerebral
Blood Flow, and Mean Tissue Transit Time, Radiology,
1999;210:519-527.
[0041] As indicated at process block 106, the next step is to
reconstruct 2D images from the acquired NMR k-space data. Each
acquired DWI image data set is transformed with a two-dimensional
complex fast Fourier transformation and the resulting I and Q
values at each image pixel are employed to calculate the magnitude
at the corresponding voxel according to the above equation (4). Six
magnitude images S.sub.1(b) . . . S.sub.6(b) are thus produced. The
magnitude of each voxel in the reconstructed reference image
S.sub.0 is also calculated.
[0042] The PWI image frames are also reconstructed using a
two-dimensional complex fast Fourier transformation. Magnitude
images are produced from the resulting I and Q values at each image
pixel according to the above equation (4). The magnitude of the NMR
signal at each voxel is thus calculated for the 46 image frames.
The resulting 46 values in the time course data at each voxel
indicates the change in NMR signal magnitude during the first pass
of the contrast agent.
[0043] As indicated at process block 108 the DWI and PWI parameter
images are calculated next. An apparent diffusion coefficient (ADC)
image is calculated with the DWI phase difference images. As is
well known in the art, this is done by calculating the diffusion
coefficient (D) for each of the six motion encoding directions
using the magnitude images S.sub.1(b)-S.sub.6(b), the reference
magnitude S.sub.0 and the above equation (1). From these six
diffusion coefficients D.sub.1-D.sub.6 the ADC is calculated.
[0044] A number of parameter images are calculated from the PWI
time course image frames. For each image pixel a concentration vs.
time curve is calculated first from the time course NMR signal
magnitude values. Integrating the concentration curve over time
yields a cerebral blood volume (CBV) value at each image pixel.
Cerebral blood flow (CBF) is then computed using deconvolution
techniques, and from the central volume theorem, the mean transit
time (MTT) is then calculated at each voxel: MTT=CBV/CBF. (6) The
DWI and PWI parameters for each voxel form a vector x for use in
the predicted outcome calculation below.
[0045] The next step in the process as indicated at process block
110 is to produce a location-weighted map which provides one or
more location parameter inputs r for each voxel. A detailed
description of the preferred method for producing the
location-weighted map is described in detail below with reference
to FIG. 4. The location weighted map provides the location
parameter value(s) at each voxel that form the vector r in the
predicted outcome calculations below.
[0046] Referring still to FIG. 3, after all the input parameter
vectors x and r are calculated, a loop is entered at 112 in which
the outcome (P) of the tissue at each voxel is predicted. As
indicated at process block 114, the x and r parameter values for
one voxel are input to the GLM model expressed above in equation
(3) and the prediction (P) is calculated for that voxel. This step
is repeated for each voxel in the examined region until predictions
for every voxel have been calculated as indicated at decision block
116. These predictions (P) are displayed as an image in which each
prediction value P color codes its corresponding image pixel. This
image is displayed as indicated at process block 118, and if a
pixel is selected by moving a cursor and "clicking" a mouse button,
the numeric prediction value (P) at the selected location is
displayed. Commands may also be entered to select voxels with a
prediction value P greater than a specified value, or between
specified values. The voxels selected in this manner are
highlighted on the image to facilitate the evaluation of clinical
outcome.
[0047] In tests conducted with and without the inclusion of this
spatial information it was found that a substantial and significant
improvement in predictive ability is achieved with the present
invention. The DWI and PWI only GLM resulted in an area under the
curve ("AUC") of 0.75.+-.0.13 (mean.+-.SD across patients) whereas
the inclusion of spatial information increased the mean AUC to
0.83.+-.0.14, a significant improvement (p=0.00096). An example of
this improved sensitivity is shown in FIG. 5 where the AUC 120 for
location+DWI+PWI weighting is 0.846 and the AUC of the DWI+PWI
weighting curve 122 is 0.712. The sensitivity of the
location+DWI+PWI weighting exceeds that of the DWI+PWI only
weighting at all false positive ratios. For the clinically
acceptable range of false positive ratios (i.e., 0 to 0.3,
corresponding to specificity ranging from 100 to 70%), the MRI-only
weighted GLM according to the present invention had a maximum
sensitivity of 68.+-.22% (which occurred at a false positive ratio
of 0.3, specificity=70%), whereas the Spatial plus MRI weighted GLM
had a maximum sensitivity of 79.+-.23%, a significant improvement
(p=0.0029, n=75).
[0048] Referring now to FIG. 4, the location weighted map which
provides the inputs to the enhanced GLM model can be produced in a
number of different ways. In the preferred embodiment a single
location parameter is produced for each voxel that indicates the
distance of the voxel from the core region 1 shown in FIG. 2. The
first step, therefore, is to identify the core region 1 as
indicated at process block 150. This can be done automatically by
selecting voxels from the smaller of the PWI acute region or the
DWI acute region. The "acute region" is those voxels having either
PWI or DWI values that indicate current cell death due to ischemia.
In the alternative, a core region 1 can be manually selected. Where
no acute region is found, which occurs in about 5% of cases, the
location of the centroid of the MTT lesion is selected as the core
region.
[0049] As indicated at process block 152, a linear distance
parameter is calculated for each voxel in the larger region 2. This
parameter (L) is the Euclidean distance in three dimensions from
the voxel location (V.sub.x,y,z) to the nearest voxel
(V.sub.x.sub.c, V.sub.y.sub.c, V.sub.z.sub.c) in the core region 1:
L= {square root over
((V.sub.x-V.sub.x.sub.c).sup.2+(V.sub.y-V.sub.y.sub.c).sup.2+(V.sub.z-V.s-
ub.z.sub.c).sup.2)}. (7) This linear distance parameter L is stored
in a 3D array at a location that corresponds to the location of the
voxel.
[0050] Other location parameters may also be calculated for the
voxel as indicated by process block 154. For example, the brain may
be segmented into regions that have different responses to the
ischemic cascade initiated by a stroke. If the voxel lies in one
region of the brain, for example, it is assigned a higher value
location parameter than a voxel located in another region of the
brain.
[0051] Possible variations on the preferred embodiment is to
replace the linear distance parameter L with one that is measured
by the distance squared from the core region 1 or by one over the
distance from the core region 1.
[0052] After the location parameters have been calculated for all
the voxels as determined at decision block 156, the resulting
location-weighted map is returned for inclusion in the predicted
outcome calculation discussed above.
[0053] The modified GLM of equation (3) used to practice the
preferred embodiment of this invention includes weighting
coefficients .beta. and .gamma. that must be determined from
training data acquired from a pool of previous patients where the
outcomes are known. As described in the above-cited publication and
co-pending U.S. patent application Ser. No. 10/182,978, this
includes selecting training regions in follow-up exams of a stroke
patient population and manually selecting regions in T2 weighted
images that clearly depict infarcted and noninfarcted tissues. The
values from DWI and PWI parameter images as well as the
location-weighted map from these same patients were used as the
input vectors x and r in the training process. The coefficients
(.beta.) and (.gamma.) are then calculated using an iterative
reweighted least-squares algorithm.
* * * * *